This project is an experimental implementation of a novel artificial general intelligence architecture based on a concept of "pressure" propagating through a field of interconnected nodes. The simulation is built with performance in mind, leveraging vectorized computations and GPU acceleration via PyTorch and the Metal Performance Shaders (MPS) backend on Apple Silicon.

The first phase of this project focused on establishing a stable, observable, and performant core simulation. Key achievements include:

• Core Data Structures: Implementation of Node and Field classes to represent the fundamental components of the system.
• GPU Acceleration: The simulation was ported from a naive O(N²) loop to a vectorized PyTorch implementation, enabling massive speed-ups on Apple Silicon via the MPS backend.
• CLI Oscilloscope: A rich-powered CLI tool was developed to visualize the state of the simulation in real-time.
• Metrics & Testing: A "coherence" metric was established to quantify system stability, with corresponding pytest tests to ensure model integrity.
• Decision Read-out: A simple decision-making function was implemented to demonstrate how the system's state can be interpreted as an output.

At the conclusion of Phase 0, a go/no-go review was conducted to determine if the project was ready to proceed to the next stage. The review was based on two key metrics:

• Coherence: The system needed to achieve a coherence score of ≥ 0.85 on a 100-node simulation.
• Decision Latency: The decision read-out latency needed to be ≤ 250 ms on the available hardware.

Decision: Green-light for Phase 1. The project successfully met all criteria and is now ready for the next phase of development.

This project is optimized for Apple Silicon (M1/M2/M3) and requires Python 3.11 or newer.

Clone the repository:

git clone https://github.com/SwiftAkira/KAI.git
cd KAI

Create a virtual environment:

python3 -m venv venv
source venv/bin/activate

Install dependencies:

pip install --upgrade pip
pip install -e .
pip install -r requirements.txt

To watch the simulation run in your terminal, use the oscilloscope.py demo. This will display a live table of node states and the current system decision.

Run on CPU:

python -m pressure_agi.demos.oscilloscope

Run on GPU (Apple Silicon):

python -m pressure_agi.demos.oscilloscope --gpu

You can also customize the number of nodes and simulation steps:

python -m pressure_agi.demos.oscilloscope --nodes 50 --steps 1000

To verify the stability and correctness of the simulation, run the test suite using pytest:

pytest

├── docs/                 # Project documentation and reports
├── pressure_agi/         # Core source code
│   ├── demos/            # Demonstrations and visualizations
│   ├── engine/           # Core simulation engine
│   └── ...
├── tests/                # Test suite
├── setup.py              # Project installation script
└── requirements.txt      # Project dependencies